The migration from IPv4 to IPv6 has been a recurring topic for years with industry reps intermittently sounding the alarm of the imminent and potentially catastrophic depletion of IPv4 addresses. The prediction that IPv4 addresses would run out by 2005 (as some analysts claimed) clearly was just alarmist. Here we are in 2009 and the Internet world has not, like Henny Penny’s sky, come crashing down. Because of network address translation (NAT), which can translate public addresses to private addresses, IPv4 longevity has been extended. However, this extension is not indefinite, and, no doubts, the transition from IPv4 to the adoption of IPv6 needs to move forward. But as with all things technological, there must be transition. This is especially pertinent given the mission-critical nature of networks and the rapid growth of traffic, services (cloud hosting for one), and products that demand increasing amounts of bandwidth.
Why should service providers make the transition from IPv4 to IPv6? IPv6 offers multiple advantages: it removes the requirement for an infinite number of addresses, offers better renumbering, facilitates multihoming, performs internal topology hiding, prevents host counting, and eliminates NAT, which all add costs and complexity to a network. These advantages make it easier for SPs to deploy new applications without having to employ difficult work-arounds or deal with random failures that are prevalent with IPv4 and NAT. Another major and challenging issue is speed and scalability. Can NAT truly translate network to network fast enough so that there is no compromise in the reliability of connection?
In addition to promoting seamless connectivity and better scaling of networks and services, IPv6 provides service provisioning, network management efficiencies, and OpEx savings. If these are not reasons enough for conversion, consider competitiveness. IPv6 not only promotes business continuity, but also fosters tech innovation and growth. For equipment manufacturers, IPv6 advances the development and evolution of smart grids, intelligent buildings, sensor networks, and other hardware- and application-dependent technologies.
Spurring innovation is another key reason for adopting this protocol. Virtually unlimited (wink) address space is expected to drive cutting-edge advances in monitoring, tracking, and remote management software and applications. What innovation adds up to is potential dollars and market share for SPs; with an adaptable platform service providers can more easily deliver new and richer services and products.
Although IPv6 is the logical next step in the evolution of the Internet, IPv6 is not a panacea. Service providers do face limitations and costs with transitioning. During incremental deployment, IPv6 features cannot be effectively and fully utilized. Additionally, because IPv6 is not a scaled up version of IPv4, and the two protocols do not interoperate directly, some form of NAT is required to translate IP packet headers between them.
Another issue to consider with IPv6 is address distribution. And here, I’ll give a word of warning: initially, it was thought that IPv4 addresses would never be exhausted, but that has not proven to be the case. When devising an IPv6 address distribution system the industry needs to very prudent in its allocation so we are not in the same boat 5 to 10 years from now. If you don’t think that this can happen just look at the growth over the last five years of cell phones with IP addresses, not to mention the many other devices that will have IP addresses.
In spite of these transitional and other technical issues, IPv6 does once fully implemented offer service providers business and economic opportunities. The question is when will the stakeholders stop crying wolf and start coordinating an industry response to IPv6 transition?
Thursday, July 9, 2009
Subscribe to:
Post Comments (Atom)
TCP/IP Class
Perception or Reality - Which Vendor(s) do you feel are more environmentally aware?
As a Service Provider how important is Green Equipment to you
4 comments:
Don't worry about IPv6 depletion (famous last words?). The standard allocation block size is /48 (2^80 addresses). This allows 65K subnets each of which is 4B times the entire IPv4 address space, for every customer. Each block is sufficient for the largest imaginable customer. There are enough such blocks already allocated (only 1/8 of the entire address space) to provide over 4,000 such blocks for every person alive on Earth today. The number of cellphones expected over the next century could be accomodated by a single /48 block if need be. even with one address for each RFID tag in existence there is no possibility of exhaustion. Ditto sensors and controls (e.g. smart buildings). Perhaps if nanomachines someday include network support, we might eat up a significant number of IPv6 addresses.
NAT has an extremely high "cost" in terms of limiting applications to either "one way" (client to server), or incorporating "NAT traversal" (a la SKYPE), which causes numerous security and complexity issues. NAT was only ever a crutch to extend the lifetime of IPv4 while IPv6 was readied. GOOD NEWS, EVERYONE! IPv6 is ready for prime time! We can throw away the (NAT) crutches. Our legs are no longer broken. We worked with Tony Hain (IPv6 guru at Cisco) to create a very good model of IPv4 exhaustion. We came up with end of 2010 for end of addresses at IANA, and six months later (mid 2011) for end of addresses at the last RIR (ARIN, APNIC, etc). This assumes no "panic allocations" or "bank runs" as the end approaches, and continuation of the current allocation rates and policies. That's right around the corner. Many groups (e.g. .AU) are not planning to be migrated until well after 2015. This is setting themselves up for disaster. This is not crying "wolf". It is OK to cry "fire" if the house really is burning down.
Hi Lawrence,
Thanks for your comments, I love it-especially the crying fire part :-)
one thing we could chat further about is NAT, I think the problem here is that your talking about technology in it's purest nature and I think a big part of it here could be what I call the 9 layers of the OSI model (politics and Religion) being layers 8 & 9---which I perceive to see that this will extent the life of NAT despite it's issues, or maybe the traditional NAT that we talk about will go away but to the SPs it becomes a vital part of the transition just needs to have some highly scalable 4 to 6 without compromise translations. We tend to forget that despite IP/NGN being the global language Frame Relay/ATM equipment is still a $2 billion a year business-because of those 8 and 9 layers and maybe an extra layer 10 called cost.
I don't know Tony Hain from Cisco-but it sounds like I should have a conversation with him, the person I hear great things about as being the main guru or fellow is Jeff Doyle from Jeff Doyle and Associates, I will reach out to him to get his latest feedback and see if I could chat with Tony from Cisco
again, thanks for your post, and I would like to have a conversation, it's always valuable for me to hear from companies like you that are out there dealing with the latest issues my e-mail is rmota@srgresearch.com
Hi Ray,
One thing I would caution against is using the predictions of analysts from more than about three years ago. I often hear people say something to the effect of, "Some people were predicting IPv4 depletion by 2007, and that didn't happen. There's no reason to believe the current predictions either."
Address usage data before 2000 was skewed somewhat by the effects of Classless Inter-Domain Routing (CIDR) policies that went into effect in 1993 and effectively slowed address usage until 2000. After 2000 IPv4 allocation took off again.
The first in-depth studies of current IPv4 allocation trends were done circa 2005 by Tony Hain of Cisco (mentioned above) and Geoff Huston of APNIC. Those first studies were wildly divergent: Tony's study predicted depletion at the IANA level around 2008, while Geoff's predicted depletion around 2021. But as both gentlemen refined their analyses and as more allocation data became available, the conclusions of the two studies (both of which are ongoing) began moving together. Both now converge on the IANA pool being exhausted around mid-2011. The five Regional Internet Registries will go dry about a year after that.
The real driver for IPv6 remains the address space -- as Lawrence Hughes points out above, an astronomically larger space than IPv4. And you point to many of the business drivers: Mobile Internet access, industrial and medical sensors, growth of the Internet into developing countries, and myriad IP-enabled applications. The overall public trend is toward a common transport technology: "everything over IP." This is fueled also by network operators consolidating divergent, service-specific networks onto a single IP network to reduce both CAPEX and OPEX (particularly the latter).
What this all boils down to is that while the characteristics of IPv6 will, once it is widely deployed, spur application innovation and simplify many aspects of network operations, for now it is a simple infrastructure issue. Service providers -- particularly broadband and mobile providers -- must have a reliable supply of new IP addresses just as they need a steady supply of new bandwidth, more powerful routers, and more versatile servers. In just a few years the only source of new IP addresses is going to be IPv6. Providers who are not preparing now for that eventuality are going to be unable to grow their business.
I want to emphasize, as many others do, that when the very last IPv4 address is handed out nothing is going to stop working. The only thing that stops is the growth of networks that cannot immediately utilize IPv6 addresses.
Post a Comment